Indirect Treatment Comparisons Using Simulation
J. Jaime Caro, United BioSource Corporation 
*K. Jack Ishak, United BioSource Corporation 

Keywords: indirect treatment comparisons, mixed treatment comparisons, simulation

In the absence of head to head trials, indirect comparison of treatments plays a central role in decision-making. Mixed treatment comparisons (MTC) have become the standard approach. Indirect comparisons are derived by linking treatment effects through common comparators based on published evidence. MTCs rely on an assumption of homogeneity between studies; that is, the effects observed in each study can be assumed to be applicable to the populations of other studies. Variability across studies is often assumed to be random and captured with mixed-effects models, which produce a summary estimate representing an average effect. Thus, the effect in any particular study – e.g., the one being considered by decision-makers – cannot be predicted specifically. We describe an approach that estimates treatment differences within the framework of a specific trial (the index trial) of a new treatment by simulating additional arms. The method aims to measure the relative benefits between the new treatment and its competitors had the latter been included in the index trial. This is done by using predictive equations for relevant clinical endpoints derived from the index trial. External data (from publications or patient-level data) describing the occurrence of the endpoints for competing treatments is used as a reference for competing treatments. A new term is added to the predictive equations to reflect comparisons with each of the alternative treatments. Endpoints are simulated from the equation for competing treatments using different potential values for the terms until one is found that yields predictions that match to reference results (from the external data source) most closely. Differences in the populations of the two trials are handled by adjusting predictions in the previous step to the characteristics of the studies of competing treatments. The relative strengths and weaknesses of the approach will be discussed as well as possible extensions.